Researchers enhance AlphaFold3 to predict protein conformational changes

Researchers at Japan’s Institute for Molecular Science (IMS) and the Graduate University for Advanced Studies, SOKENDAI, have enhanced the predictive capabilities of AlphaFold3, the leading artificial‑intelligence system for protein structure modeling. By introducing a repulsive force between generated conformations, the team enabled the algorithm to explore a broader range of structural states, addressing a long‑standing limitation in which AlphaFold3’s default settings typically converge on a single, static configuration despite proteins often adopting multiple functional forms.

The modified approach allows AlphaFold3 to sample diverse conformational landscapes, offering a more realistic representation of protein dynamics that underpins many biological processes. This advancement could improve computational studies of enzyme mechanisms, drug target validation, and the design of biomolecules, providing researchers with a more versatile tool for investigating the structural flexibility essential to protein function.